Power-User · Level 3 of 3
The Home Assistant Voice Preview Edition (Voice PE) is a $69 smart speaker that processes voice commands for your smart home without sending them to a company’s servers, provided you already run Home Assistant. It controls lights, thermostats, timers, and shopping lists reliably, and its physical mute switch and open firmware make it genuinely private. It does not stream Spotify, run third-party skills, or match the polish of an Amazon Echo out of the box, and independent reviewers have flagged inconsistent recognition on anything outside its trained phrasing. If you already run Home Assistant and want a private way to talk to it, it earns its price. If you want a plug-and-play Alexa replacement, wait.
What the Voice Preview Edition Actually Is
The Voice PE is a satellite microphone and speaker, not a smart home hub on its own. Nabu Casa, the company behind Home Assistant Cloud, designed it to sit alongside an existing Home Assistant installation — something like Home Assistant Green, a Raspberry Pi running Home Assistant OS, or a spare mini PC already running Home Assistant. Without one of those, the Voice PE has nothing to talk to.
Physically, it is a small white puck built around an ESP32-S3 chip. Two microphones and an XMOS audio processor pick your voice out of background noise, a rotating dial adjusts volume with a click, and the middle button doubles as a programmable smart button for triggering automations with a double, triple, or long press. A physical mute switch cuts power to the microphones (the digital mute is separate), and a 3.5mm jack with its own DAC lets you route audio to a real speaker. It runs on USB-C power (no charger included) and connects over 2.4GHz Wi-Fi and Bluetooth 5.0. Firmware, hardware schematics, and software are all open source.
What It Costs, Realistically
Home Assistant lists the Voice PE at a recommended $69 (€59) MSRP, up from its $59 launch price when it shipped in December 2024. That number only tells part of the story, since the device does nothing without a Home Assistant instance behind it.
| Starting Point | Approx. Total Cost | Notes |
|---|---|---|
| You already run Home Assistant (Pi, NAS, old laptop, mini PC) | $69 | Just the Voice PE itself |
| Starting from zero with Home Assistant Green | $69 + $199 = $268 | Plug-and-play hub, no separate computer to manage |
| Add Home Assistant Cloud (optional) | + $6.50/mo or $65/yr | Widest language support, easiest cloud voice processing, Alexa/Google bridging |
Home Assistant Green’s price rose from $99 at launch to $199 in 2026; Nabu Casa attributes the increase directly to rising RAM component costs and says it has tried to avoid raising the price again soon. If $268 to get started feels steep, Home Assistant’s own installation guide points out that a computer you already own — an old laptop, a spare Raspberry Pi, a NAS — can run Home Assistant well enough to skip the Green purchase entirely.
Choosing How Voice Processing Runs: Local or Cloud
The single most important setup decision with the Voice PE is where speech gets processed. Home Assistant gives you three options, and picking the right one depends on your hardware and your language.
- Focused Local (Speech-to-Phrase). Runs entirely on low-power hardware and recognizes a defined set of home-control phrases — “turn on the light,” “set a timer for 5 minutes.” It cannot process open-ended speech: Home Assistant’s own documentation gives the example that it can turn on a device but cannot add an item to your shopping list. Choose this if your hub is modest (a Raspberry Pi, Home Assistant Green) and your commands are mostly device control.
- Full Local (Whisper). Processes general speech entirely on your own hardware, with no phrase list to memorize. Home Assistant recommends at least an Intel N100-equivalent processor for acceptable speed; on weaker hardware, responses can take several seconds. Choose this if you have that hardware already and want zero cloud dependency for anything you say.
- Home Assistant Cloud. Offloads speech processing to Microsoft Azure’s enterprise speech service under a no-retention agreement, giving low-powered hardware cloud-grade accuracy and the widest language coverage. Choose this if your language isn’t well supported locally yet, or if you would rather pay $6.50 a month than manage local speech models.
None of the three requires a subscription to use the Voice PE at all — the cloud tier is an option, not a gate. That is the real pitch: choice, not a forced monthly fee for basic function.
What It Handles Well
- Core device control. Lights, switches, thermostats, fans, and media players you have exposed to Assist respond reliably to direct commands like “turn on the light” or “turn on the fan.”
- Timers and lists. “Set a timer for 5 minutes” and “add milk to my shopping list” are built-in, documented commands that work without extra configuration (outside the Focused Local model’s limits above).
- Real privacy controls. The physical mute switch cuts power to the microphones, not just software access to them — a meaningfully different guarantee than a software toggle.
- A smart button you already have. The middle button supports single, double, triple, and long presses as automation triggers, and the LED ring itself is a controllable RGB light you can use in automations (a visual medication reminder, for instance).
- Room to tinker. Open firmware, an exposed Grove port, and accessible PCB pads mean the hardware itself is meant to be modified, not just used.
Where It Falls Short of Alexa
Home Assistant is direct about this in its own FAQ: the Voice PE cannot yet do everything an Echo, Google Nest, or HomePod can, and the company frames it as a preview rather than a finished competitor.
- Music streaming is not native. There is no built-in Spotify or Apple Music skill. The internal speaker is meant for spoken feedback, not listening, and Home Assistant’s own guidance is to add a Music Assistant server plus an external speaker over the 3.5mm jack if you want real audio.
- Only three wake words. Out of the box you can pick “Okay Nabu,” “Hey Jarvis,” or “Hey Mycroft.” Custom wake words require training data and hardware most users do not have; there is no “Alexa” or “computer” option today.
- No skill store. Alexa’s value for a lot of households is its library of third-party skills. The Voice PE has no equivalent; anything beyond home control and utility commands requires connecting a cloud or local LLM, which Home Assistant itself has called experimental and recommends using cautiously for anything important.
- Recognition friction outside trained phrases. A hands-on review from Smart Home Solver, published in February 2025, described inconsistent results with anything beyond direct commands: skip-song requests worked unreliably, resuming playback required saying “resume” rather than the more natural “play,” and unrecognized requests fell back to a generic “sorry, I couldn’t understand that.” That was one reviewer’s experience over a year before this article was written, on a product Home Assistant continues to actively develop, so treat it as a snapshot rather than a permanent verdict.
Who Should Wait
Skip the Voice PE for now if you want one device that plays Spotify, answers trivia, and controls your smart home without any setup — that is still Alexa or Google Assistant’s territory. It is also not the right starting point if you have no interest in running Home Assistant itself; the speaker is an accessory to that system, not a replacement for it. And if your language is not yet supported by the local speech models, you will end up leaning on Home Assistant Cloud anyway, which narrows the cost gap with a subscription-based assistant.
It is a good fit if you already run Home Assistant on some kind of hub, care about keeping voice data out of a corporate cloud, and are comfortable troubleshooting the occasional misheard command while the project matures.
Why This Is Getting More Attention Right Now
Part of the renewed interest in local voice assistants traces back to Amazon’s own product. Alexa+, Amazon’s AI-upgraded assistant, is free for Prime members and costs $20 a month for everyone else, and Amazon has been auto-upgrading existing Echo owners onto it on a rolling schedule rather than waiting for people to opt in. That has generated real frustration among long-time Echo owners who did not ask for the change, and it is part of why a locally processed, no-subscription-required device like the Voice PE is getting a second look from people who would not have considered a hobbyist smart speaker two years ago. Worth being clear-eyed about the trade: you are swapping a mature, broad assistant for a private, narrower one that is still explicitly labeled a preview.
If you are weighing a full switch away from Alexa rather than just adding a private satellite mic, the Alexa vs. Google Home vs. Apple HomeKit comparison is a better starting point — it covers ecosystem-level trade-offs the Voice PE alone does not touch.
Pricing and specs above come directly from Home Assistant’s and Nabu Casa’s own product and pricing pages, checked July 2026. The hands-on impressions of everyday voice recognition come from one published third-party review from February 2025 and are identified as such — this article was written from published documentation and reporting, not from testing the device directly.
The short version: the Voice PE is a well-built, honestly marketed preview of what a private voice assistant could be, priced fairly for what it does today. It is worth buying if you already have a Home Assistant setup to plug it into and value privacy over polish. It is not yet worth buying as your only smart speaker.